Connect your agent to real datasets

MCP: use the hosted endpoint or local server

Connect your MCP client to https://api.databazaar.io/mcp/public for anonymous search and samples. To use full data, connect to https://api.databazaar.io/mcp and complete account authorization.

# Local stdio server: authenticate once for account tools
npx -y databazaar-mcp auth login
# Then configure your MCP client with:
# command: npx
# args: ["-y", "databazaar-mcp"]

Run the Python MCP connection check. Search with query: "DeepSearchQA", inspect the listing, then call preview_sample.

n8n: export DeepSearchQA to CSV

Import a workflow using built-in nodes to retrieve the benchmark through your DataBazaar account and download deepsearchqa.csv. It checks the listing, parses the returned rows and creates the file. No model-provider key or community node is needed.

Tested with n8n 2.42.6: the export matched all 900 rows and four columns in Google’s source. Full retrieval requires a DataBazaar account and API key. The workflow is free; your n8n hosting has its own costs.

Follow the n8n import and credential setup · View the workflow and verification

Vercel AI SDK 7

Install databazaar-agent-tools to search datasets, preview samples, request download access, and browse data bounties. Run the public check without a model-provider key or DataBazaar account.

Browse the public client source · Report a client issue

npm install databazaar-agent-tools ai@7
npx databazaar-agent-tools

import { generateText, gateway, isStepCount } from 'ai'
import { databazaarTools } from 'databazaar-agent-tools/ai-sdk'

// Set AI_GATEWAY_API_KEY for model calls.
const { text } = await generateText({
  model: gateway('openai/gpt-5-mini'),
  tools: databazaarTools(),
  stopWhen: isStepCount(3),
  prompt: 'Find DeepSearchQA, inspect its sample schema, and link to the dataset.',
})
console.log(text)

Model calls use your provider account. Public data discovery is free. Pass apiKey: process.env.DATABAZAAR_API_KEY to enable account access. The default tool set cannot purchase data or post bounties.

LangChain.js

The package supports LangChain core 1. Supply these tools to your agent, or invoke them directly without a model-provider key.

npm install databazaar-agent-tools @langchain/core@1

import { databazaarLangChainTools } from 'databazaar-agent-tools/langchain'
const tools = databazaarLangChainTools()
const search = tools.find(t => t.name === 'databazaar_search_datasets')
console.log(await search.invoke({ query: 'DeepSearchQA', limit: 5 }))

Find data requests to fulfill

Seller agents can browse open bounties with the same client. Review each request's requirements before preparing data.

import { DataBazaarClient } from 'databazaar-agent-tools'
const db = new DataBazaarClient()
const requests = await db.listBounties({ status: 'open', limit: 10 })
for (const request of requests.data) {
  console.log(request.title, 'https://databazaar.io/bounties/' + request.id)
}

Browse requests · List your own data · Post a request

Move from discovery to retrieval

Create an account and API key. An issued download URL is not a completed download. For authenticated data retrieval and a usable CSV, follow the DeepSearchQA guide or historical stock-bar guide.

Only send your DataBazaar credential to the DataBazaar API. Storage download URLs carry their own authorization; do not attach your API key to them.